Diagnostic Utility of Bronchoscopy in Adults with Acute Myeloid Leukemia and Other High-Grade Myeloid Neoplasms

LETTER TO THE EDITOR

Respiratory symptoms and chest radiological abnormalities are common in adults with acute myeloid leukemia (AML) and other high-grade myeloid neoplasms [1]. For evaluation, flexible fiberoptic bronchoscopy with bronchioalveolar lavage (BAL) is frequently used, primarily supported by limited data from patients treated over a decade ago [2–7]. With new antimicrobials (e.g. broad-spectrum azoles [8]) and non-invasive diagnostic tools (e.g. quantifying serum galactomannan [GM] [9]) available, the utility of BAL may be reduced. Similarly, although typically considered safe, the incidence of clinically significant procedural adverse events is unknown in this population – uncertainties prompting our retrospective analysis of the utility and risks of BAL at our institution between June 2008 and December 2015.

Via electronic medical records, we identified adults with high-grade myeloid neoplasms who underwent diagnostic BAL at any point during their cancer treatment. Chart abstraction was performed by two independent physicians, with a third adjudicating differences; abstracted data included microbiology findings, final diagnoses, “bronchoscopy-only diagnoses” that could not be made through blood or sputum testing; “unexpected diagnoses” that were not entertained in any pre-procedure physician note; and “leading diagnoses” posited in the pulmonologist’s pre-procedure note. We also recorded how often results changed clinical management (addition/discontinuation of treatment(s), procedural interventions, or use of results to justify additional AML therapy). We recorded all procedural complications and considered as serious: respiratory failure requiring mechanical ventilation, pulmonary hemorrhage, hypotension requiring vasopressors, pneumothorax requiring chest tube, or other life-threatening events. Linear regression analyses tested associated between patient- and disease-related covariates (Supplementary Table 1) and bronchoscopic utility in three pre-defined stringency levels (unexpected diagnosis with management change, discordant diagnosis with management change, any management change). Generalized estimating equation methodology accounted for repeated observations. Analyses were performed with Stata version 14 (StataCorp; College Station, TX, USA).

During the 7.5-year period, 391 of the identified 1,442 patients underwent 461 bronchoscopies (1: n=330; 2: n=50; 3: n=7; 4: n=1; 6: n=1). Of these, 59% were performed in patients who were not in remission at the time of BAL. Almost two thirds had grade 4 neutropenia and 24% received systemic immunosuppressive therapy (IST) at the time of BAL. IST included standard post-transplant suppression alone (n=52) or in combination with corticosteroids (n=59). Almost 75% of bronchoscopies were performed to assess pulmonary symptoms, while the remainder evaluated fever (61%) or asymptomatic radiographic abnormalities (39%); for patient characteristics, see Supplementary Table 2. There was heterogeneity in which diagnostic tests were performed prior to and during bronchoscopy (Supplementary Figure 1). Invasive testing (from bronchoscopy) with high yield included GM (148/454 [33%]), fungal PCR (53/219 [24%]), aspergillus PCR (68/457 [15%]), viral PCR (63/421 [15%]), bacterial culture (54/458 [12%]), and zygomyces PCR (11/120 [9%]). Bronchoscopy established a diagnosis in 62% (287/461), half of which (114/287) were “bronchoscopy-only” diagnoses established solely through invasive testing (Figure 1). Additionally, infection with multiple organisms was detected in 25 cases (5.4%). Complications occurred in 15% of cases (n=71), but serious complications were infrequent, occurring in 1.1% of cases (n=5: pneumothorax requiring chest tube, hypotension requiring vasopressors, life-threatening arrhythmia, hypoxemia requiring mechanical ventilation, and stroke); no patient had more than one serious complication.

Figure 1.

Diagnostic findings from each of the 460 completed bronchoscopies.

The “true” clinical value of bronchoscopy is difficult to assess in retrospective analyses because a change in therapy after bronchoscopy does not necessarily reflect a change due to bronchoscopy. Had bronchoscopy been unavailable, how often would similar changes have been made empirically? To address this challenge, we investigated in how many cases 1) an unexpected diagnosis was made and followed by a management change (as the most rigorous estimate of utility), 2) the post-bronchoscopy diagnosis was discordant from the leading diagnosis considered before this procedure and was followed by a management change, and 3) a change in management was made following bronchoscopy regardless of whether the diagnosis was expected or considered.

Bronchoscopy lead to “unexpected” diagnoses in 66 cases (Table 1), of which 61 resulted in a change in management (13% of total): infections (n=51 including Pneumocystis jirovecii, Influenza, Legionella, Toxoplasma, invasive mold), inflammatory conditions (n=14 including diffuse alveolar hemorrhage [DAH], sarcoidosis, pulmonary alveolar proteinosis, eosinophilic pneumonia, amyloidosis), and chloroma (n=1), most of which are not treated by standard empiric therapies.

TABLE 1.

Unexpected diagnoses (n=66) and paired leading clinical diagnoses.

Bacterial pneumonia (n=8)
Staphylococcal pneumonia (4)	Thought to be: fungal pneumonia in all cases
Legionella pneumonia (2)	Thought to be: fungal pneumonia in all cases
Actinomyces pneumonia (2)	Thought to be: fungal pneumonia in all cases
Fungal pneumonia (n=21)
Invasive mold (11)	Thought to be: other fungal infection (10) or lung cancer (1)
Candida pneumonia (3)	Thought to be: other infectious pneumonia in all cases
Pneumocystis pneumonia (3)	Thought to be: other infectious pneumonia in all cases
Malessezia pneumonia (2)	Thought to be: aspergillus pneumonia or ARDS
Blastomyces + Cryptococcal pneumonia (1)	Thought to be: idiopathic pulmonary syndrome
Geotricum pneumonia (1)	Thought to be: bacterial pneumonia
Viral pneumonia (n=13)	Thought to be: other infectious pneumonia (10), diffuse alveolar hemorrhage (2), non-influenza viral pneumonia (1)
Mixed infection pneumonia (n=8)	Thought to be: single infection pneumonia in all cases
Parasitic pneumonia (n=1)
Toxoplasmosis (1)	Thought to be: other infectious pneumonia
Inflammatory conditions (n=14)
Diffuse alveolar hemorrhage (7)	Thought to be: infectious pneumonia (6), ARDS (1)
Cryptogenic organizing PNA. (2)	Thought to be: infectious pneumonia in all cases
Idiopathic pulmonary syndrome (1)	Thought to be: infectious pneumonia
Amyloidosis (1)	Thought to be: infectious pneumonia
Sarcoidosis (1)	Thought to be: infectious pneumonia
Eosinophilic pneumonia (1)	Thought to be: infectious pneumonia
Pulmonary alveolar proteinosis (1)	Thought to be: infectious pneumonia
Leukemia / chloroma (n=1)	Thought to be: other infectious pneumonia

It is plausible the utility of bronchoscopy/BAL exceeded the 13% of cases in which an unforeseen diagnosis changed management. There are likely cases in which a diagnosis is entertained, but starting empiric therapy would not be considered without confirmation because of competing diagnostic possibilities and/or the potential for significant treatment side effects. In our series, the final bronchoscopic diagnosis was discordant from the leading pre-procedure diagnosis in 213/443 (48%) of cases, of which 116 resulted in change in management. Another definition of bronchoscopic utility would therefore be 26% (116/443).

When approaching the concept of “utility” as any procedure leading to a management change, we found medical care was modified in 266 (58%) of cases, with addition (30%, n=139) and/or de-escalation (32%, n=147) of antimicrobials, addition of steroids and supportive measures for inflammatory conditions (9%, n=42), procedures such as lung resection for invasive mold (2%, n=7), or clearing patients to receive AML therapy (2%, n=11). The fact that bronchoscopy altered management in a majority of cases may largely reflect the broad definition of “change” used in our study, as we counted any alteration, not just addition of antimicrobials. However, as a contributor, the yield of infectious organisms (54% in our study) was higher than in earlier studies (36–48%) [3–5], possibly due to increased use of molecular testing.

Multivariable analysis highlighted several populations in which bronchoscopy may have increased utility (Supplementary Table 1). Bronchoscopies in female patients more likely yielded unexpected diagnoses resulting in change in antimicrobials, even though the prevalence of major diagnostic categories did not vary by gender. We can only speculate whether this finding is explained by “atypical” disease presentations, differential reporting of symptoms, or differences in how physicians interpret symptoms in women. Further studies to examine gender disparities in the utility of bronchoscopy may be warranted. If we consider “utility” more broadly as any case in which a discordant diagnosis was followed by change in management, hypoxemia requiring supplemental oxygen as a presenting sign was most associated with utility. This finding is driven largely by an association with inflammatory conditions (e.g., DAH), as these were found in 20% of patients with hypoxemia and only 3% of patients without (P<0.0001) and prompted addition of steroids and other supportive measures. When “utility” is defined as any change in management, different factors become associated with utility: in particular, cough and consolidation were associated with diagnosis of fungal infection, which allowed for more informed antibiotic selection. Notably, patients receiving IST were not more likely to have useful bronchoscopy results than the other patients on multivariable analysis, regardless of how utility is defined.

Additionally, there were factors associated with decreased bronchoscopic yield. Patients with positive serum GM (31/302 checked) were unlikely to have any additional findings on bronchoscopy: in only 4 of the 31 cases (all with serum GM just above the threshold of 0.5 [range 0.53–0.65]) did BAL reveal additional diagnoses. The use of posaconazole prophylaxis, which increased after 2015, was associated with fewer diagnoses of fungal pneumonia (20% vs. 40%, P=0.02) and Aspergillus in particular (7% vs. 32%, P<0.001) but not the rates of other infections. Notably, however, two-thirds of patients diagnosed with Aspergillus (91/136) were on effective anti-fungal therapy at the time of bronchoscopic diagnosis; thus, these diagnostic studies remain relevant in treated patients. Another factor that could have decreased the yield of bronchoscopy was that biopsies were rarely performed (9 biopsies total in 8 out of 461 procedures), likely due to severe thrombocytopenia in the study cohort (median platelet count 38,000/μL [inter-quartile range: 25,000–62,000/μL]). As some diagnoses relevant for post-transplant patients require biopsy, lack of biopsy could limit the diagnostic capability of bronchoscopy.

Although limited by its retrospective nature, ours is the largest study of bronchoscopic utility in patients with myeloid neoplasms. Our results suggest the benefits of bronchoscopy/BAL may outweigh the risks and may have clinical utility in this patient population. The difference between the percentage of cases where the procedure can reasonably be considered “useful” (13–26%) and the percentage leading to serious complications (1%) supports this claim. While bronchoscopy appears to alter treatment, it remains unknown whether bronchoscopy improves outcomes. Two prior studies that compared survival in patients in whom bronchoscopy did vs. did not yield a microbiological diagnosis reached differing conclusions [2,3]. Such retrospective comparisons, however, are limited by challenges in controlling for measured covariates, to say nothing of the unmeasured and unknown. Retrospective chart reviews such as ours cannot collect outcomes for patients in whom bronchoscopy was considered but not performed, and therefore conclusions about the performance of bronchoscopy in all-comers with AML must be interpreted cautiously. While a prospective randomized study of bronchoscopy as a diagnostic modality would be ideal, such a study may never be conducted. In this setting, our data provides an argument for the utility of bronchoscopy in evaluating/managing patients with high-grade myeloid neoplasms who present with respiratory signs or symptoms.